Disclosed herein is a machine learning application for automatically reading filled-in forms. There are multiple steps involved in using a computer to accurately read a handwritten form. First, the system identifies the form. Second, the system identifies what parts of the form are important. Third, the important parts are extracted as image data (known as shreds). Finally, fourth, the system interprets the shreds. This application is focused on steps two and three of that overall process. The disclosed techniques relate to training a machine learning system on a given series of forms such that when provided future filled-in forms within that series, the system is able to extract the portions of the filled-in form that are important/relevant.

A computer-implemented method of generating electronic documents is described. The method comprises receiving a plurality of scanned documents for a plurality of vehicles; providing the plurality of scanned documents to a neural network model that outputs respective class identifiers of the plurality of scanned documents; for each scanned document of the plurality of scanned documents, extracting data from the scanned document according to a corresponding class identifier, and associating the scanned document and the extracted data with an identified vehicle of the plurality of vehicles, wherein the identified vehicle is identified by the extracted data; and generating an electronic document for the identified vehicle using the extracted data.

A system and method for deconstructing a document is described herein, where the method is an improvement over existing document deconstruction techniques. These improvements increase speed and accuracy by rapidly identifying the source in a document by splitting the document into a plurality of sections and performing a perceptual image hashing on each section. Then a hamming distance is used to compare the hash for each section with the hashes of known documents to identify the source who sent the document.

Aspects of the disclosure provide for mechanisms for identification of objects in images using neural networks. A method of the disclosure includes: obtaining an image, representing each element of a plurality of elements of the image via an input vector of a plurality of input vectors, each input vector having one or more parameters pertaining to visual appearance of a respective element of the image, providing the plurality of input vectors to a first subnetwork of a neural network to obtain a plurality of output vectors, wherein each of the plurality of output vectors is associated with an element of the image, identifying, based on the plurality of output vectors, a sub-plurality of elements of the image as belonging to the image of the object, and determining, based on locations of the sub-plurality of elements, a location of an image of an object within the image.

A verification apparatus according to one embodiment of the present disclosure selects an image associated with a code of a reference image switching sheet as a reference image when the code of the reference image switching sheet is read. An instruction is provided to discharge the reference image switching sheet to a set sheet discharging destination.

Image quality metrics are applied to the entire image within the field of view of a camera associated with a mobile device and are combined with non-image metrics applied to the device to provide a hybrid approach to capturing a quality image that is not dependent upon the check image satisfying a monitoring criterion. When the composite image passes image monitoring criteria and the device passes a non-image criterion, an image may be captured by the camera either manually or automatically. The process may be applied to capture images of both the front and the back of the check, and may be applied to multiple checks. The images may then be submitted for processing for deposit of the check or checks into the user's bank account without the requirement that the user review and approve the image or images of the back of the check or checks.

An information processing device is provided with an acquisition unit, a detection unit, and a combination unit. The acquisition unit performs layout analysis on image information and acquires multiple regions. The detection unit detects a feature indicating that regions are continuous from each of the multiple regions acquired by the acquisition unit. The combination unit combines adjacent regions in a case in which the feature is detected by the detection unit.

A field-of-view of a scene is scanned by an augmented reality device. The scene includes one or more objects including a first computing device. A portion of an electronic document is detected based on the scanned field-of-view. The portion of the electronic document is rendered on a display of the first computing device. A content element of the electronic document that is rendered on the display is captured. A second computing device determines an incomplete portion of the content element. A suggestion to complete the incomplete portion is provided by the augmented reality device.

A collation device includes a processor configured to, by executing a program: (a) acquire a photographed image including a collation area provided on a printing substrate having unevenness, (b) simultaneously execute smoothing processing and shading difference enhancement processing on the photographed image, and (c) detect the collation area based on an image obtained by simultaneously executing the smoothing processing and the shading difference enhancement processing.

A handwritten content removing method and device and a storage medium. The handwritten content removing method comprises: acquiring an input image of a text page to be processed, the input image comprising a handwritten region, which comprises a handwritten content (S10); identifying the input image so as to determine the handwritten content in the handwritten region (S11); and removing the handwritten content in the input image so as to obtain an output image (S12).